Originally posted by wormwoodThere is implied criticism here and I can't understand why. Chess Tempo can be used for free and you pay for some extra stuff if you want it. Isn't that the same as this site?
...I popped on your site. and just like I suspected, you've made it commercial now. 35$ a year for a CTS ri ting on these things. I thought you were better than that. my mistake.
follow the money.
I'll provide an extreme example
White to move.
In a puzzle, I would play 1. Kf1! (it wins) as there is no difference between 1. dxe3 and 1. fxe3.
ChessTempo's interface however would point out to me that 1. fxe3 mates in 11 and 1. dxe3 only in 12. But I would not even have tried 1. dxe3 - because it reaches the same outcome as the other move. 😉
Originally posted by heinzkatthat's very extreme
I'll provide an extreme example
[fen]4k3/8/8/8/8/4r3/3P1P2/3RK3 w - -[/fen]
White to move.
In a puzzle, I would play 1. Kf1! as there is no difference between 1. dxe3 and 1. fxe3.
ChessTempo's interface however would point out to me that 1. fxe3 mates in 11 and 1. dxe3 only in 12. But I would not even have tried 1. dxe3 - because it reaches the same outcome as the other move. 😉
Originally posted by heinzkatThe guy who used to do TrainingBot and ProblemBot on ICC had a better solution for situations like this one: just give the user credit for both moves that are easily winning.
I'll provide an extreme example
[fen]4k3/8/8/8/8/4r3/3P1P2/3RK3 w - -[/fen]
White to move.
In a puzzle, I would play 1. Kf1! as there is no difference between 1. dxe3 and 1. fxe3.
ChessTempo's interface however would point out to me that 1. fxe3 mates in 11 and 1. dxe3 only in 12. But I would not even have tried 1. dxe3 - because it reaches the same outcome as the other move. 😉
Even in puzzles that start off with two-three 'clear best' moves, they can reach a point in which there are two equally good continuations. Rather than throw that puzzle out just because it's not dual-free, it's better [from a tactical training standpoint] to give credit for both, or just end the problem at that point.
Originally posted by SwissGambitJust end the problem before that point is the way it "should" be done, to prevent silly solvers like Heinzkat from being confused. But on CT, often on the first move there are already alternatives. I'd say throw all of those out, and we start talking, but others do not quite agree 😉
Just give the user credit for both moves that are easily winning.
Even in puzzles that start off with two-three 'clear best' moves, they can reach a point in which there are two equally good continuations. Rather than throw that puzzle out just because it's not dual-free, it's better [from a tactical training standpoint] to give credit for both, or just end the problem at that point.
Originally posted by heinzkatEven in your Qb7/Qb8 example, if one leads to +5 and the other to +7, I say give credit for both moves. No need to throw the problem out.
Just end the problem before that point is the way it "should" be done, to prevent silly solvers like Heinzkat from being confused. But on CT, often on the first move there are already alternatives. I'd say throw all of those out, and we start talking, but others do not quite agree 😉
To help the solver, the program can point out that the 'other' move also wins. That way it's clear that some puzzles have more than one 'right' move. If the goal is to find ideas, rather than just 'winning-est winning move', this should be OK.
Originally posted by SwissGambitI skip two moves that reach the same outcome, much like solving a PG (2. Nb5 is never possible) or h#
Even in your Qb7/Qb8 example, if one leads to +5 and the other to +7, I say give credit for both moves. No need to throw the problem out.
Originally posted by heinzkatWell, that's great for PGs, and compositions in general, but in tactical training, I fail to see why the positions must have a dual-free solution. If anything, in chess games, there are often positions with two or more equally good continuations.
I skip two moves that reach the same outcome, much like solving a PG (2. Nb5 is never possible) or h#
Hi heinzkat,
That's an interesting example, in that situation you'd risk failure if that position was on CT as so far, the tactics generator doesn't use tablebases so without plugging that into an engine I'd be surprised if an engine could find the win with 1.Kf1. I'm probably missing something in the discussion, but I'm surprised you'd prefer a mate in 42 over a mate in 11 and 12, but it does make a reasonable point. Actually I just let the engine have a look at this without tablebases, it does find the two shorter mates, but the 1.Kf1 gets evaled high enough to be given a 'try again', so you'd at least get a chance to have another try.
In terms of making those who don't like the 'alts' system and the 'try agains' a bit happier, the most likely change will be to simply introduced a rated mode where none of the problems have alts, I'm not quite at the point where I have enough problems for this, but I'm getting there.
SwissGambit: I don't like cutting problems short when you reach an ambiguity point. I'd prefer to throw the problem out if this was the only option. The reason is, that it's fine for high rated players who can fill in the gaps, but for lower rated users they are often left scracthing their heads when a problem finished before the final tactical point has been made. Allowing both lines is tempting but also has some issues, it works ok on trainingbot because it's unrated, but there are issues that can make it hard to implement this type of thing fairly in a rated environment. One of the main ICC training bot problem editors is a prolific CT user and has participated in many discussions on this topic, he seems reasonably comfortable with the current alts approach (although again I acknowledge that not everyone is!).
Regards,
Richard.
Originally posted by richardchesstempoLook, this is what is fundamentally wrong about CT, a lot of talking and dwelling that is completely and solely supported by engines; "what I say can't be wrong, look, the engine gives it +2.50!" without any further input from the problems/commenters themselves. I didn't give the example to be looked at by your engine. I gave the example for a review by YOU. 😲
Actually I just let the engine have a look at this (yeah I ripped it out of context)
Originally posted by heinzkatI think that is an extremely unkind interpretation of what I said. I'm not sure which part of what I said you've interpreted as saying 'what I say can't be wrong, look the engine gives it +2.50'. In fact I was saying exactly the opposite and pointing out the limitations of computer analysis, especially in the type of position you provided where the computer can't actually see that your move wins. I said 'I'd be surprised if an engine could find the win with 1.Kf1.', and then I ended up putting it into the engine to verify that it would indeed NOT find it, and it didn't (although it decided the eval was high enough for the move not to be punished). Please explain which part of my statement was saying 'what I say can't be wrong, look the engine gives it +2.50'. Having said that, if you're doing a position designed to be used for practical training purposes and you choose a move that takes you 31 moves longer to mate then I think it's reasonable that you are told 'your move wasn't good enough'. Generally speaking I agree that 'a mate, is a mate, is a mate', but there are practical time management (and in some cases 50 move rule) issues that need to be considered as well, which means there are limitations to 'any mate wins, it shouldn't matter which one I play'.
Look, this is what is fundamentally wrong about CT, a lot of talking and dwelling that is completely and solely supported by engines; "what I say can't be wrong, look, the engine gives it +2.50!" without any further input from the problems/commenters themselves. I didn't give the example to be looked at by your engine. I gave the example for a review by YOU. 😲
You'll be extremely disappointed if you expect the CT positions to behave like composed problems, as generally speaking they don't. It's probably a mistake for me to call them 'problems' as this can perhaps create this expectation. As training positions they are mostly effective (judged by the feedback of hundreds of people far more skilled than I am), there are some degenerative cases, but even the examples from the site you've had issues with I think can be argued to have some training value, although I wouldn't say they are shining examples.
Regards,
Richard.